1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
5 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
6 <link rel="stylesheet" href="llvm.css" type="text/css">
7 <title>LLVM 3.0 Release Notes</title>
11 <h1>LLVM 3.0 Release Notes</h1>
13 <img align=right src="http://llvm.org/img/DragonSmall.png"
14 width="136" height="136" alt="LLVM Dragon Logo">
17 <li><a href="#intro">Introduction</a></li>
18 <li><a href="#subproj">Sub-project Status Update</a></li>
19 <li><a href="#externalproj">External Projects Using LLVM 3.0</a></li>
20 <li><a href="#whatsnew">What's New in LLVM 3.0?</a></li>
21 <li><a href="GettingStarted.html">Installation Instructions</a></li>
22 <li><a href="#knownproblems">Known Problems</a></li>
23 <li><a href="#additionalinfo">Additional Information</a></li>
26 <div class="doc_author">
27 <p>Written by the <a href="http://llvm.org/">LLVM Team</a></p>
31 <h1 style="color:red">These are in-progress notes for the upcoming LLVM 3.0
34 <a href="http://llvm.org/releases/2.9/docs/ReleaseNotes.html">LLVM 2.9
35 Release Notes</a>.</h1>
38 <!-- *********************************************************************** -->
40 <a name="intro">Introduction</a>
42 <!-- *********************************************************************** -->
46 <p>This document contains the release notes for the LLVM Compiler
47 Infrastructure, release 3.0. Here we describe the status of LLVM, including
48 major improvements from the previous release, improvements in various
49 subprojects of LLVM, and some of the current users of the code.
50 All LLVM releases may be downloaded from
51 the <a href="http://llvm.org/releases/">LLVM releases web site</a>.</p>
53 <p>For more information about LLVM, including information about the latest
54 release, please check out the <a href="http://llvm.org/">main LLVM web
55 site</a>. If you have questions or comments,
56 the <a href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVM
57 Developer's Mailing List</a> is a good place to send them.</p>
59 <p>Note that if you are reading this file from a Subversion checkout or the main
60 LLVM web page, this document applies to the <i>next</i> release, not the
61 current one. To see the release notes for a specific release, please see the
62 <a href="http://llvm.org/releases/">releases page</a>.</p>
67 <!-- *********************************************************************** -->
69 <a name="subproj">Sub-project Status Update</a>
71 <!-- *********************************************************************** -->
75 <p>The LLVM 3.0 distribution currently consists of code from the core LLVM
76 repository (which roughly includes the LLVM optimizers, code generators and
77 supporting tools), and the Clang repository. In
78 addition to this code, the LLVM Project includes other sub-projects that are
79 in development. Here we include updates on these subprojects.</p>
81 <!--=========================================================================-->
83 <a name="clang">Clang: C/C++/Objective-C Frontend Toolkit</a>
88 <p><a href="http://clang.llvm.org/">Clang</a> is an LLVM front end for the C,
89 C++, and Objective-C languages. Clang aims to provide a better user
90 experience through expressive diagnostics, a high level of conformance to
91 language standards, fast compilation, and low memory use. Like LLVM, Clang
92 provides a modular, library-based architecture that makes it suitable for
93 creating or integrating with other development tools. Clang is considered a
94 production-quality compiler for C, Objective-C, C++ and Objective-C++ on x86
95 (32- and 64-bit), and for Darwin/ARM targets.</p>
97 <p>In the LLVM 3.0 time-frame, the Clang team has made many improvements:
99 <li>Greatly improved support for building C++ applications, with greater
100 stability and better diagnostics.</li>
102 <li><a href="http://clang.llvm.org/cxx_status.html">Improved support</a> for
103 the <a href="http://www.iso.org/iso/iso_catalogue/catalogue_tc/catalogue_detail.htm?csnumber=50372">C++
104 2011</a> standard (aka "C++'0x"), including implementations of non-static data member
105 initializers, alias templates, delegating constructors, range-based
106 for loops, and implicitly-generated move constructors and move assignment
107 operators, among others.</li>
109 <li>Implemented support for some features of the upcoming C1x standard,
110 including static assertions and generic selections.</li>
112 <li>Better detection of include and linking paths for system headers and
113 libraries, especially for Linux distributions.</li>
115 <li>Several improvements to Objective-C support, including:
118 <li><a href="http://clang.llvm.org/docs/AutomaticReferenceCounting.html">
119 Automatic Reference Counting</a> (ARC) and an improved memory model
120 cleanly separating object and C memory.</li>
122 <li>A migration tool for moving manual retain/release code to ARC</li>
124 <li>Better support for data hiding, allowing instance variables to be
125 declared in implementation contexts or class extensions</li>
126 <li>Weak linking support for Objective-C classes</li>
127 <li>Improved static type checking by inferring the return type of methods
128 such as +alloc and -init.</li>
131 Some new Objective-C features require either the Mac OS X 10.7 / iOS 5
132 Objective-C runtime, or version 1.6 or later of the GNUstep Objective-C
133 runtime version.</li>
135 <li>Implemented a number of optimizations in <tt>libclang</tt>, the Clang C
136 interface, to improve the performance of code completion and the mapping
137 from source locations to abstract syntax tree nodes.</li>
139 For more details about the changes to Clang since the 2.9 release, see the
140 <a href="http://clang.llvm.org/docs/ReleaseNotes.html">Clang release notes</a>
144 <p>If Clang rejects your code but another compiler accepts it, please take a
145 look at the <a href="http://clang.llvm.org/compatibility.html">language
146 compatibility</a> guide to make sure this is not intentional or a known
151 <!--=========================================================================-->
153 <a name="dragonegg">DragonEgg: GCC front-ends, LLVM back-end</a>
157 <p><a href="http://dragonegg.llvm.org/">DragonEgg</a> is a
158 <a href="http://gcc.gnu.org/wiki/plugins">gcc plugin</a> that replaces GCC's
159 optimizers and code generators with LLVM's. It works with gcc-4.5 or gcc-4.6,
160 targets the x86-32 and x86-64 processor families, and has been successfully
161 used on the Darwin, FreeBSD, KFreeBSD, Linux and OpenBSD platforms. It fully
162 supports Ada, C, C++ and Fortran. It has partial support for Go, Java, Obj-C
165 <p>The 3.0 release has the following notable changes:</p>
168 <li>GCC version 4.6 is now fully supported.</li>
170 <li>Patching and building GCC is no longer required: the plugin should work
171 with your system GCC (version 4.5 or 4.6; on Debian/Ubuntu systems the
172 gcc-4.5-plugin-dev or gcc-4.6-plugin-dev package is also needed).</li>
174 <li>The <tt>-fplugin-arg-dragonegg-enable-gcc-optzns</tt> option, which runs
175 GCC's optimizers as well as LLVM's, now works much better. This is the
176 option to use if you want ultimate performance! It is still experimental
177 though: it may cause the plugin to crash.</li>
179 <li>The type and constant conversion logic has been almost entirely rewritten,
180 fixing a multitude of obscure bugs.</li>
186 <!--=========================================================================-->
188 <a name="compiler-rt">compiler-rt: Compiler Runtime Library</a>
193 <p>The new LLVM <a href="http://compiler-rt.llvm.org/">compiler-rt project</a>
194 is a simple library that provides an implementation of the low-level
195 target-specific hooks required by code generation and other runtime
196 components. For example, when compiling for a 32-bit target, converting a
197 double to a 64-bit unsigned integer is compiled into a runtime call to the
198 "__fixunsdfdi" function. The compiler-rt library provides highly optimized
199 implementations of this and other low-level routines (some are 3x faster than
200 the equivalent libgcc routines).</p>
202 <p>In the LLVM 3.0 timeframe, the target specific ARM code has converted to
203 "unified" assembly syntax, and several new functions have been added to the
208 <!--=========================================================================-->
210 <a name="lldb">LLDB: Low Level Debugger</a>
215 <p>LLDB is a ground-up implementation of a command line debugger, as well as a
216 debugger API that can be used from other applications. LLDB makes use of the
217 Clang parser to provide high-fidelity expression parsing (particularly for
218 C++) and uses the LLVM JIT for target support.</p>
220 <p>LLDB has advanced by leaps and bounds in the 3.0 timeframe. It is
221 dramatically more stable and useful, and includes both a
222 new <a href="http://lldb.llvm.org/tutorial.html">tutorial</a> and
223 a <a href="http://lldb.llvm.org/lldb-gdb.html">side-by-side comparison with
228 <!--=========================================================================-->
230 <a name="libc++">libc++: C++ Standard Library</a>
235 <p>Like compiler_rt, libc++ is now <a href="DeveloperPolicy.html#license">dual
236 licensed</a> under the MIT and UIUC license, allowing it to be used more
239 <p>Libc++ has been ported to FreeBSD and imported into the base system. It is
240 planned to be the default STL implementation for FreeBSD 10.</p>
244 <!--=========================================================================-->
246 <a name="vmkit">VMKit</a>
251 <p>The <a href="http://vmkit.llvm.org/">VMKit project</a> is an
252 implementation of a Java Virtual Machine (Java VM or JVM) that uses LLVM for
253 static and just-in-time compilation.
255 <p>In the LLVM 3.0 time-frame, VMKit has had significant improvements on both
256 runtime and startup performance:</p>
259 <li>Precompilation: by compiling ahead of time a small subset of Java's core
260 library, the startup performance have been highly optimized to the point that
261 running a 'Hello World' program takes less than 30 milliseconds.</li>
263 <li>Customization: by customizing virtual methods for individual classes,
264 the VM can statically determine the target of a virtual call, and decide to
267 <li>Inlining: the VM does more inlining than it did before, by allowing more
268 bytecode instructions to be inlined, and thanks to customization. It also
269 inlines GC barriers, and object allocations.</li>
271 <li>New exception model: the generated code for a method that does not do
272 any try/catch is not penalized anymore by the eventuality of calling a
273 method that throws an exception. Instead, the method that throws the
274 exception jumps directly to the method that could catch it.</li>
280 <!--=========================================================================-->
282 <a name="LLBrowse">LLBrowse: IR Browser</a>
287 <p><a href="http://llvm.org/svn/llvm-project/llbrowse/trunk/doc/LLBrowse.html">
288 LLBrowse</a> is an interactive viewer for LLVM modules. It can load any LLVM
289 module and displays its contents as an expandable tree view, facilitating an
290 easy way to inspect types, functions, global variables, or metadata nodes. It
291 is fully cross-platform, being based on the popular wxWidgets GUI
297 <!--=========================================================================-->
300 <a name="klee">KLEE: A Symbolic Execution Virtual Machine</a>
305 <a href="http://klee.llvm.org/">KLEE</a> is a symbolic execution framework for
306 programs in LLVM bitcode form. KLEE tries to symbolically evaluate "all" paths
307 through the application and records state transitions that lead to fault
308 states. This allows it to construct testcases that lead to faults and can even
309 be used to verify some algorithms.
317 <!-- *********************************************************************** -->
319 <a name="externalproj">External Open Source Projects Using LLVM 3.0</a>
321 <!-- *********************************************************************** -->
325 <p>An exciting aspect of LLVM is that it is used as an enabling technology for
326 a lot of other language and tools projects. This section lists some of the
327 projects that have already been updated to work with LLVM 3.0.</p>
329 <!--=========================================================================-->
330 <h3>AddressSanitizer</h3>
334 <p><a href="http://code.google.com/p/address-sanitizer/">AddressSanitizer</a>
335 uses compiler instrumentation and a specialized malloc library to find C/C++
336 bugs such as use-after-free and out-of-bound accesses to heap, stack, and
337 globals. The key feature of the tool is speed: the average slowdown
338 introduced by AddressSanitizer is less than 2x.</p>
342 <!--=========================================================================-->
347 <p><a href="http://www.clamav.net">Clam AntiVirus</a> is an open source (GPL)
348 anti-virus toolkit for UNIX, designed especially for e-mail scanning on mail
351 <p>Since version 0.96 it
352 has <a href="http://vrt-sourcefire.blogspot.com/2010/09/introduction-to-clamavs-low-level.html">bytecode
353 signatures</a> that allow writing detections for complex malware.
354 It uses LLVM's JIT to speed up the execution of bytecode on X86, X86-64,
355 PPC32/64, falling back to its own interpreter otherwise. The git version was
356 updated to work with LLVM 3.0.</p>
360 <!--=========================================================================-->
361 <h3>clang_complete for VIM</h3>
365 <p><a href="https://github.com/Rip-Rip/clang_complete">clang_complete</a> is a
366 VIM plugin, that provides accurate C/C++ autocompletion using the clang front
367 end. The development version of clang complete, can directly use libclang
368 which can maintain a cache to speed up auto completion.</p>
372 <!--=========================================================================-->
377 <p><a href="https://bitbucket.org/dwilliamson/clreflect">clReflect</a> is a C++
378 parser that uses clang/LLVM to derive a light-weight reflection database
379 suitable for use in game development. It comes with a very simple runtime
380 library for loading and querying the database, requiring no external
381 dependencies (including CRT), and an additional utility library for object
382 management and serialisation.</p>
386 <!--=========================================================================-->
387 <h3>Cling C++ Interpreter</h3>
391 <p><a href="http://cern.ch/cling">Cling</a> is an interactive compiler interface
392 (aka C++ interpreter). It supports C++ and C, and uses LLVM's JIT and the
393 Clang parser. It has a prompt interface, runs source files, calls into shared
394 libraries, prints the value of expressions, even does runtime lookup of
395 identifiers (dynamic scopes). And it just behaves like one would expect from
400 <!--=========================================================================-->
401 <h3>Crack Programming Language</h3>
405 <p><a href="http://code.google.com/p/crack-language/">Crack</a> aims to provide
406 the ease of development of a scripting language with the performance of a
407 compiled language. The language derives concepts from C++, Java and Python,
408 incorporating object-oriented programming, operator overloading and strong
413 <!--=========================================================================-->
418 <p><a href="http://eerolanguage.org/">Eero</a> is a fully
419 header-and-binary-compatible dialect of Objective-C 2.0, implemented with a
420 patched version of the Clang/LLVM compiler. It features a streamlined syntax,
421 Python-like indentation, and new operators, for improved readability and
422 reduced code clutter. It also has new features such as limited forms of
423 operator overloading and namespaces, and strict (type-and-operator-safe)
424 enumerations. It is inspired by languages such as Smalltalk, Python, and
429 <!--=========================================================================-->
430 <h3>FAUST Real-Time Audio Signal Processing Language</h3>
434 <p><a href="http://faust.grame.fr/">FAUST</a> is a compiled language for
435 real-time audio signal processing. The name FAUST stands for Functional
436 AUdio STream. Its programming model combines two approaches: functional
437 programming and block diagram composition. In addition with the C, C++, Java
438 output formats, the Faust compiler can now generate LLVM bitcode, and works
444 <!--=========================================================================-->
445 <h3>Glasgow Haskell Compiler (GHC)</h3>
449 <p>GHC is an open source, state-of-the-art programming suite for Haskell, a
450 standard lazy functional programming language. It includes an optimizing
451 static compiler generating good code for a variety of platforms, together
452 with an interactive system for convenient, quick development.</p>
454 <p>GHC 7.0 and onwards include an LLVM code generator, supporting LLVM 2.8 and
455 later. Since LLVM 2.9, GHC now includes experimental support for the ARM
456 platform with LLVM 3.0.</p>
460 <!--=========================================================================-->
465 <p><a href="http://botwars.tk/gwscript/">gwXscript</a> is an object oriented,
466 aspect oriented programming language which can create both executables (ELF,
467 EXE) and shared libraries (DLL, SO, DYNLIB). The compiler is implemented in
468 its own language and translates scripts into LLVM-IR which can be optimized
469 and translated into native code by the LLVM framework. Source code in
470 gwScript contains definitions that expand the namespaces. So you can build
471 your project and simply 'plug out' features by removing a file. The remaining
472 project does not leave scars since you directly separate concerns by the
473 'template' feature of gwX. It is also possible to add new features to a
474 project by just adding files and without editing the original project. This
475 language is used for example to create games or content management systems
476 that should be extendable.</p>
478 <p>gwXscript is strongly typed and offers comfort with its native types string,
479 hash and array. You can easily write new libraries in gwXscript or native
480 code. gwXscript is type safe and users should not be able to crash your
481 program or execute malicious code except code that is eating CPU time.</p>
485 <!--=========================================================================-->
486 <h3>include-what-you-use</h3>
490 <p><a href="http://code.google.com/p/include-what-you-use">include-what-you-use</a>
491 is a tool to ensure that a file directly <code>#include</code>s
492 all <code>.h</code> files that provide a symbol that the file uses. It also
493 removes superfluous <code>#include</code>s from source files.</p>
497 <!--=========================================================================-->
498 <h3>ispc: The Intel SPMD Program Compiler</h3>
502 <p><a href="http://ispc.github.com">ispc</a> is a compiler for "single program,
503 multiple data" (SPMD) programs. It compiles a C-based SPMD programming
504 language to run on the SIMD units of CPUs; it often delivers 5-6x speedups on
505 a single core of a CPU with an 8-wide SIMD unit compared to serial code,
506 while still providing a clean and easy-to-understand programming model. For
507 an introduction to the language and its performance,
508 see <a href="http://ispc.github.com/example.html">the walkthrough</a> of a short
509 example program. ispc is licensed under the BSD license.</p>
513 <!--=========================================================================-->
514 <h3>The Julia Programming Language</h3>
518 <p><a href="http://github.com/JuliaLang/julia">Julia</a> is a high-level,
519 high-performance dynamic language for technical
520 computing. It provides a sophisticated compiler, distributed parallel
521 execution, numerical accuracy, and an extensive mathematical function
522 library. The compiler uses type inference to generate fast code
523 without any type declarations, and uses LLVM's optimization passes and
524 JIT compiler. The language is designed around multiple dispatch,
525 giving programs a large degree of flexibility. It is ready for use on many
526 kinds of problems.</p>
529 <!--=========================================================================-->
530 <h3>LanguageKit and Pragmatic Smalltalk</h3>
534 <p><a href="http://etoileos.com/etoile/features/languagekit/">LanguageKit</a> is
535 a framework for implementing dynamic languages sharing an object model with
536 Objective-C. It provides static and JIT compilation using LLVM along with
537 its own interpreter. Pragmatic Smalltalk is a dialect of Smalltalk, built on
538 top of LanguageKit, that interfaces directly with Objective-C, sharing the
539 same object representation and message sending behaviour. These projects are
540 developed as part of the Étoilé desktop environment.</p>
544 <!--=========================================================================-->
549 <p><a href="http://lua-av.mat.ucsb.edu/blog/">LuaAV</a> is a real-time
550 audiovisual scripting environment based around the Lua language and a
551 collection of libraries for sound, graphics, and other media protocols. LuaAV
552 uses LLVM and Clang to JIT compile efficient user-defined audio synthesis
553 routines specified in a declarative syntax.</p>
557 <!--=========================================================================-->
562 <p>An open source, cross-platform implementation of C# and the CLR that is
563 binary compatible with Microsoft.NET. Has an optional, dynamically-loaded
564 LLVM code generation backend in Mini, the JIT compiler.</p>
566 <p>Note that we use a Git mirror of LLVM <a
567 href="https://github.com/mono/llvm">with some patches</a>.</p>
571 <!--=========================================================================-->
576 <p><a href="http://polly.grosser.es">Polly</a> is an advanced data-locality
577 optimizer and automatic parallelizer. It uses an advanced, mathematical
578 model to calculate detailed data dependency information which it uses to
579 optimize the loop structure of a program. Polly can speed up sequential code
580 by improving memory locality and consequently the cache use. Furthermore,
581 Polly is able to expose different kind of parallelism which it exploits by
582 introducing (basic) OpenMP and SIMD code. A mid-term goal of Polly is to
583 automatically create optimized GPU code.</p>
587 <!--=========================================================================-->
588 <h3>Portable OpenCL (pocl)</h3>
592 <p>Portable OpenCL is an open source implementation of the OpenCL standard which
593 can be easily adapted for new targets. One of the goals of the project is
594 improving performance portability of OpenCL programs, avoiding the need for
595 target-dependent manual optimizations. A "native" target is included, which
596 allows running OpenCL kernels on the host (CPU).</p>
600 <!--=========================================================================-->
604 <p><a href="http://pure-lang.googlecode.com/">Pure</a> is an
605 algebraic/functional programming language based on term rewriting. Programs
606 are collections of equations which are used to evaluate expressions in a
607 symbolic fashion. The interpreter uses LLVM as a backend to JIT-compile Pure
608 programs to fast native code. Pure offers dynamic typing, eager and lazy
609 evaluation, lexical closures, a hygienic macro system (also based on term
610 rewriting), built-in list and matrix support (including list and matrix
611 comprehensions) and an easy-to-use interface to C and other programming
612 languages (including the ability to load LLVM bitcode modules, and inline C,
613 C++, Fortran and Faust code in Pure programs if the corresponding LLVM-enabled
614 compilers are installed).</p>
616 <p>Pure version 0.48 has been tested and is known to work with LLVM 3.0
617 (and continues to work with older LLVM releases >= 2.5).</p>
621 <!--=========================================================================-->
622 <h3>Renderscript</h3>
626 <p><a href="http://developer.android.com/guide/topics/renderscript/index.html">Renderscript</a>
627 is Android's advanced 3D graphics rendering and compute API. It provides a
628 portable C99-based language with extensions to facilitate common use cases
629 for enhancing graphics and thread level parallelism. The Renderscript
630 compiler frontend is based on Clang/LLVM. It emits a portable bitcode format
631 for the actual compiled script code, as well as reflects a Java interface for
632 developers to control the execution of the compiled bitcode. Executable
633 machine code is then generated from this bitcode by an LLVM backend on the
634 device. Renderscript is thus able to provide a mechanism by which Android
635 developers can improve performance of their applications while retaining
640 <!--=========================================================================-->
645 <p><a href="http://safecode.cs.illinois.edu">SAFECode</a> is a memory safe C/C++
646 compiler built using LLVM. It takes standard, unannotated C/C++ code,
647 analyzes the code to ensure that memory accesses and array indexing
648 operations are safe, and instruments the code with run-time checks when
649 safety cannot be proven statically. SAFECode can be used as a debugging aid
650 (like Valgrind) to find and repair memory safety bugs. It can also be used
651 to protect code from security attacks at run-time.</p>
655 <!--=========================================================================-->
656 <h3>The Stupid D Compiler (SDC)</h3>
660 <p><a href="https://github.com/bhelyer/SDC">The Stupid D Compiler</a> is a
661 project seeking to write a self-hosting compiler for the D programming
662 language without using the frontend of the reference compiler (DMD).</p>
666 <!--=========================================================================-->
667 <h3>TTA-based Co-design Environment (TCE)</h3>
671 <p>TCE is a toolset for designing application-specific processors (ASP) based on
672 the Transport triggered architecture (TTA). The toolset provides a complete
673 co-design flow from C/C++ programs down to synthesizable VHDL and parallel
674 program binaries. Processor customization points include the register files,
675 function units, supported operations, and the interconnection network.</p>
677 <p>TCE uses Clang and LLVM for C/C++ language support, target independent
678 optimizations and also for parts of code generation. It generates new
679 LLVM-based code generators "on the fly" for the designed TTA processors and
680 loads them in to the compiler backend as runtime libraries to avoid
681 per-target recompilation of larger parts of the compiler chain.</p>
685 <!--=========================================================================-->
686 <h3>Tart Programming Language</h3>
690 <p><a href="http://code.google.com/p/tart/">Tart</a> is a general-purpose,
691 strongly typed programming language designed for application
692 developers. Strongly inspired by Python and C#, Tart focuses on practical
693 solutions for the professional software developer, while avoiding the clutter
694 and boilerplate of legacy languages like Java and C++. Although Tart is still
695 in development, the current implementation supports many features expected of
696 a modern programming language, such as garbage collection, powerful
697 bidirectional type inference, a greatly simplified syntax for template
698 metaprogramming, closures and function literals, reflection, operator
699 overloading, explicit mutability and immutability, and much more. Tart is
700 flexible enough to accommodate a broad range of programming styles and
701 philosophies, while maintaining a strong commitment to simplicity, minimalism
702 and elegance in design.</p>
706 <!--=========================================================================-->
707 <h3>ThreadSanitizer</h3>
711 <p><a href="http://code.google.com/p/data-race-test/">ThreadSanitizer</a> is a
712 data race detector for (mostly) C and C++ code, available for Linux, Mac OS
713 and Windows. On different systems, we use binary instrumentation frameworks
714 (Valgrind and Pin) as frontends that generate the program events for the race
715 detection algorithm. On Linux, there's an option of using LLVM-based
716 compile-time instrumentation.</p>
722 <!-- *********************************************************************** -->
724 <a name="whatsnew">What's New in LLVM 3.0?</a>
726 <!-- *********************************************************************** -->
730 <p>This release includes a huge number of bug fixes, performance tweaks and
731 minor improvements. Some of the major improvements and new features are
732 listed in this section.</p>
734 <!--=========================================================================-->
736 <a name="majorfeatures">Major New Features</a>
741 <!-- Features that need text if they're finished for 3.1:
745 loop dependence analysis
746 CorrelatedValuePropagation
747 lib/Transforms/IPO/MergeFunctions.cpp => consider for 3.1.
748 Integrated assembler on by default for arm/thumb?
753 Analysis/RegionInfo.h + Dom Frontiers
754 SparseBitVector: used in LiveVar.
755 llvm/lib/Archive - replace with lib object?
758 <p>LLVM 3.0 includes several major changes and big features:</p>
761 <li>llvm-gcc is no longer supported, and not included in the release. We
762 recommend switching to <a
763 href="http://clang.llvm.org/">Clang</a> or <a
764 href="http://dragonegg.llvm.org/">DragonEgg</a>.</li>
766 <li>The linear scan register allocator has been replaced with a new "greedy"
767 register allocator, enabling live range splitting and many other
768 optimizations that lead to better code quality. Please see its <a
769 href="http://blog.llvm.org/2011/09/greedy-register-allocation-in-llvm-30.html">blog post</a> or its talk at the <a
770 href="http://llvm.org/devmtg/2011-11/">Developer Meeting</a>
771 for more information.</li>
772 <li>LLVM IR now includes full support for <a href="Atomics.html">atomics
773 memory operations</a> intended to support the C++'11 and C'1x memory models.
774 This includes <a href="LangRef.html#memoryops">atomic load and store,
775 compare and exchange, and read/modify/write instructions</a> as well as a
776 full set of <a href="LangRef.html#ordering">memory ordering constraints</a>.
777 Please see the <a href="Atomics.html">Atomics Guide</a> for more
780 <li>The LLVM IR exception handling representation has been redesigned and
781 reimplemented, making it more elegant, fixing a huge number of bugs, and
782 enabling inlining and other optimizations. Please see its blog post (XXX
783 not yet) and the <a href="ExceptionHandling.html">Exception Handling
784 documentation</a> for more information.</li>
785 <li>The LLVM IR Type system has been redesigned and reimplemented, making it
786 faster and solving some long-standing problems.
788 href="http://blog.llvm.org/2011/11/llvm-30-type-system-rewrite.html">blog
789 post</a> for more information.</li>
791 <li>The MIPS backend has made major leaps in this release, going from an
792 experimental target to being virtually production quality and supporting a
793 wide variety of MIPS subtargets. See the <a href="#MIPS">MIPS section</a>
794 below for more information.</li>
796 <li>The optimizer and code generator now supports gprof and gcov-style coverage
797 and profiling information, and includes a new llvm-cov tool (but also works
798 with gcov). Clang exposes coverage and profiling through GCC-compatible
799 command line options.</li>
805 <!--=========================================================================-->
807 <a name="coreimprovements">LLVM IR and Core Improvements</a>
812 <p>LLVM IR has several new features for better support of new targets and that
813 expose new optimization opportunities:</p>
816 <li><a href="Atomics.html">Atomic memory accesses and memory ordering</a> are
817 now directly expressible in the IR.</li>
818 <li>A new <a href="LangRef.html#int_fma">llvm.fma intrinsic</a> directly
819 represents floating point multiply accumulate operations without an
820 intermediate rounding stage.</li>
821 <li>A new llvm.expect intrinsic (XXX not documented in langref) allows a
822 frontend to express expected control flow (and the __builtin_expect builtin
824 <li>The <a href="LangRef.html#int_prefetch">llvm.prefetch intrinsic</a> now
825 takes a 4th argument that specifies whether the prefetch happens from the
826 icache or dcache.</li>
827 <li>The new <a href="LangRef.html#uwtable">uwtable function attribute</a>
828 allows a frontend to control emission of unwind tables.</li>
829 <li>The new <a href="LangRef.html#fnattrs">nonlazybind function
830 attribute</a> allow optimization of Global Offset Table (GOT) accesses.</li>
831 <li>The new <a href="LangRef.html#returns_twice">returns_twice attribute</a>
832 allows better modeling of functions like setjmp.</li>
833 <li>The <a href="LangRef.html#datalayout">target datalayout</a> string can now
834 encode the natural alignment of the target's stack for better optimization.
839 <!--=========================================================================-->
841 <a name="optimizer">Optimizer Improvements</a>
846 <p>In addition to many minor performance tweaks and bug fixes, this
847 release includes a few major enhancements and additions to the
851 <li>The pass manager now has an extension API that allows front-ends and plugins
852 to insert their own optimizations in the well-known places in the standard
853 pass optimization pipeline.</li>
855 <li>Information about <a href="BranchWeightMetadata.html">branch probability</a>
856 and basic block frequency is now available within LLVM, based on a
857 combination of static branch prediction heuristics and
858 <code>__builtin_expect</code> calls. That information is currently used for
859 register spill placement and if-conversion, with additional optimizations
860 planned for future releases. The same framework is intended for eventual
861 use with profile-guided optimization.</li>
863 <li>The "-indvars" induction variable simplification pass only modifies
864 induction variables when profitable. Sign and zero extension
865 elimination, linear function test replacement, loop unrolling, and
866 other simplifications that require induction variable analysis have
867 been generalized so they no longer require loops to be rewritten into
868 canonical form prior to optimization. This new design
869 preserves more IR level information, avoids undoing earlier loop
870 optimizations (particularly hand-optimized loops), and no longer
871 requires the code generator to reconstruct loops into an optimal form -
872 an intractable problem.</li>
874 <li>LLVM now includes a pass to optimize retain/release calls for the
875 <a href="http://clang.llvm.org/docs/AutomaticReferenceCounting.html">Automatic
876 Reference Counting</a> (ARC) Objective-C language feature (in
877 lib/Transforms/Scalar/ObjCARC.cpp). It is a decent example of implementing
878 a source-language-specific optimization in LLVM.</li>
884 <!--=========================================================================-->
886 <a name="mc">MC Level Improvements</a>
891 <p>The LLVM Machine Code (aka MC) subsystem was created to solve a number of
892 problems in the realm of assembly, disassembly, object file format handling,
893 and a number of other related areas that CPU instruction-set level tools work
894 in. For more information, please see
895 the <a href="http://blog.llvm.org/2010/04/intro-to-llvm-mc-project.html">Intro
896 to the LLVM MC Project Blog Post</a>.</p>
899 <li>The MC layer has undergone significant refactoring to eliminate layering
900 violations that caused it to pull in the LLVM compiler backend code.</li>
901 <li>The ELF object file writers are much more full featured.</li>
902 <li>The integrated assembler now supports #line directives.</li>
903 <li>An early implementation of a JIT built on top of the MC framework (known
904 as MC-JIT) has been implemented and will eventually replace the old JIT.
905 It emits object files direct to memory and uses a runtime dynamic linker to
906 resolve references and drive lazy compilation. The MC-JIT enables much
907 greater code reuse between the JIT and the static compiler and provides
908 better integration with the platform ABI as a result.
910 <li>The assembly printer now makes uses of assemblers instruction aliases
911 (InstAliases) to print simplified mneumonics when possible.</li>
912 <li>TableGen can now autogenerate MC expansion logic for pseudo
913 instructions that expand to multiple MC instructions (through the
914 PseudoInstExpansion class).</li>
915 <li>A new llvm-dwarfdump tool provides a start of a drop-in
916 replacement for the corresponding tool that use LLVM libraries. As part of
917 this, LLVM has the beginnings of a dwarf parsing library.</li>
918 <li>llvm-objdump has more output including, symbol by symbol disassembly,
919 inline relocations, section headers, symbol tables, and section contents.
920 Support for archive files has also been added.</li>
921 <li>llvm-nm has gained support for archives of binary files.</li>
922 <li>llvm-size has been added. This tool prints out section sizes.</li>
927 <!--=========================================================================-->
929 <a name="codegen">Target Independent Code Generator Improvements</a>
934 <p>We have put a significant amount of work into the code generator
935 infrastructure, which allows us to implement more aggressive algorithms and
936 make it run faster:</p>
939 <li>XXX: Segmented stacks.</li>
940 <li>LLVM generates substantially better code for indirect gotos due to a new
941 tail duplication pass, which can be a substantial performance win for
942 interpreter loops that use them.</li>
943 <li>Exception handling and debug frame information is now emitted with CFI
944 directives. This lets the assembler produce more compact info as it knows
945 the final offsets, yielding <a href="http://blog.mozilla.com/respindola/2011/05/12/cfi-directives/">much smaller executables</a> for some C++ applications.
946 If the system assembler doesn't support it, MC exands the directives when
947 the integrated assembler is not used.
950 <li>The code generator now supports vector "select" operations on vector
951 comparisons, turning them into various optimized code sequences (e.g.
952 using the SSE4/AVX "blend" instructions).</li>
953 <li>The SSE execution domain fix pass and the ARM NEON move fix pass have been
954 merged to a target independent execution dependency fix pass. This pass is
955 used to select alternative equivalent opcodes in a way that minimizes
956 execution domain crossings. Closely connected instructions are moved to
957 the same execution domain when possible. Targets can override the
958 <code>getExecutionDomain</code> and <code>setExecutionDomain</code> hooks
959 to use the pass.</li>
963 <!--=========================================================================-->
965 <a name="x86">X86-32 and X86-64 Target Improvements</a>
970 <p>New features and major changes in the X86 target include:</p>
973 <li>The X86 backend, assembler and disassembler now have full support for AVX 1.
974 To enable it pass <code>-mavx</code> to the compiler. AVX2 implementation is
975 underway on mainline.</li>
976 <li>The integrated assembler and disassembler now support a broad range of new
977 instructions including Atom, Ivy Bridge, <a
978 href="http://en.wikipedia.org/wiki/SSE4a">SSE4a/BMI</a> instructions, <a
979 href="http://en.wikipedia.org/wiki/RdRand">rdrand</a> and many others.</li>
980 <li>The X86 backend now fully supports the <a href="http://llvm.org/PR879">X87
981 floating point stack inline assembly constraints</a>.</li>
982 <li>The integrated assembler now supports the <tt>.code32</tt> and
983 <tt>.code64</tt> directives to switch between 32-bit and 64-bit
985 <li>The X86 backend now synthesizes horizontal add/sub instructions from generic
986 vector code when the appropriate instructions are enabled.</li>
987 <li>The X86-64 backend generates smaller and faster code at -O0 due to
988 improvements in fast instruction selection.</li>
989 <li><a href="http://code.google.com/p/nativeclient/">Native Client</a>
990 subtarget support has been added.</li>
992 <li>The CRC32 intrinsics have been renamed. The intrinsics were previously
993 <code>@llvm.x86.sse42.crc32.[8|16|32]</code>
994 and <code>@llvm.x86.sse42.crc64.[8|64]</code>. They have been renamed to
995 <code>@llvm.x86.sse42.crc32.32.[8|16|32]</code> and
996 <code>@llvm.x86.sse42.crc32.64.[8|64]</code>.</li>
1001 <!--=========================================================================-->
1003 <a name="ARM">ARM Target Improvements</a>
1008 <p>New features of the ARM target include:</p>
1011 <li>The ARM backend generates much faster code for Cortex-A9 chips.</li>
1012 <li>The ARM backend has improved support for Cortex-M series processors.</li>
1013 <li>The ARM inline assembly constraints have been implemented and are now fully
1015 <li>NEON code produced by Clang often runs much faster due to improvements in
1016 the Scalar Replacement of Aggregates pass.</li>
1017 <li>The old ARM disassembler is replaced with a new one based on autogenerated
1018 encoding information from ARM .td files.</li>
1019 <li>The integrated assembler has made major leaps forward, but is still beta quality in LLVM 3.0.</li>
1024 <!--=========================================================================-->
1026 <a name="MIPS">MIPS Target Improvements</a>
1031 <p>This release has seen major new work on just about every aspect of the MIPS
1032 backend. Some of the major new features include:</p>
1035 <li>Most MIPS32r1 and r2 instructions are now supported.</li>
1036 <li>LE/BE MIPS32r1/r2 has been tested extensively.</li>
1037 <li>O32 ABI has been fully tested.</li>
1038 <li>MIPS backend has migrated to using the MC infrastructure for assembly printing. Initial support for direct object code emission has been implemented too.</li>
1039 <li>Delay slot filler has been updated. Now it tries to fill delay slots with useful instructions instead of always filling them with NOPs.</li>
1040 <li>Support for old-style JIT is complete.</li>
1041 <li>Support for old architectures (MIPS1 and MIPS2) has been removed.</li>
1042 <li>Initial support for MIPS64 has been added.</li>
1046 <!--=========================================================================-->
1048 <a name="PTX">PTX Target Improvements</a>
1054 The PTX back-end is still experimental, but is fairly usable for compute kernels
1055 in LLVM 3.0. Most scalar arithmetic is implemented, as well as intrinsics to
1056 access the special PTX registers and sync instructions. The major missing
1057 pieces are texture/sampler support and some vector operations.</p>
1059 <p>That said, the backend is already being used for domain-specific languages
1060 and can be used by Clang to
1061 <a href="http://clang.llvm.org/docs/ReleaseNotes.html#opencl">compile OpenCL
1062 C code</a> into PTX.</p>
1066 <!--=========================================================================-->
1068 <a name="OtherTS">Other Target Specific Improvements</a>
1074 <li>Many PowerPC improvements have been implemented for ELF targets, including
1075 support for varargs and initial support for direct .o file emission.</li>
1077 <li>MicroBlaze scheduling itineraries were added that model the
1078 3-stage and the 5-stage pipeline architectures. The 3-stage
1079 pipeline model can be selected with <code>-mcpu=mblaze3</code>
1080 and the 5-stage pipeline model can be selected with
1081 <code>-mcpu=mblaze5</code>.</li>
1087 <!--=========================================================================-->
1089 <a name="changes">Major Changes and Removed Features</a>
1094 <p>If you're already an LLVM user or developer with out-of-tree changes based on
1095 LLVM 2.9, this section lists some "gotchas" that you may run into upgrading
1096 from the previous release.</p>
1099 <li>LLVM 3.0 removes support for reading LLVM 2.8 and earlier files, and LLVM
1100 3.1 will eliminate support for reading LLVM 2.9 files. Going forward, we
1101 aim for all future versions of LLVM to read bitcode files and .ll files
1102 produced by LLVM 3.0.</li>
1103 <li>Tablegen has been split into a library, allowing the clang tblgen pieces
1104 to now live in the clang tree. The llvm version has been renamed to
1105 llvm-tblgen instead of tblgen.</li>
1106 <li>The <code>LLVMC</code> meta compiler driver was removed.</li>
1107 <li>The unused PostOrder Dominator Frontiers and LowerSetJmp passes were removed.</li>
1110 <li>The old <code>TailDup</code> pass was not used in the standard pipeline
1111 and was unable to update ssa form, so it has been removed.
1112 <li>The syntax of volatile loads and stores in IR has been changed to
1113 "<code>load volatile</code>"/"<code>store volatile</code>". The old
1114 syntax ("<code>volatile load</code>"/"<code>volatile store</code>")
1115 is still accepted, but is now considered deprecated and will be removed in
1117 <li>llvm-gcc's frontend tests have been removed from llvm/test/Frontend*, sunk
1118 into the clang and dragonegg testsuites.</li>
1119 <li>The old atomic intrinsics (<code>llvm.memory.barrier</code> and
1120 <code>llvm.atomic.*</code>) are now gone. Please use the new atomic
1121 instructions, described in the <a href="Atomics.html">atomics guide</a>.
1122 <li>LLVM's configure script doesn't depend on llvm-gcc anymore, eliminating a
1123 strange circular dependence between projects.</li>
1126 <h4>Windows (32-bit)</h4>
1130 <li>On Win32(MinGW32 and MSVC), Windows 2000 will not be supported.
1131 Windows XP or higher is required.</li>
1138 <!--=========================================================================-->
1140 <a name="api_changes">Internal API Changes</a>
1145 <p>In addition, many APIs have changed in this release. Some of the major
1146 LLVM API changes are:</p>
1149 <li>The biggest and most pervasive change is that the type system has been
1150 rewritten: <code>PATypeHolder</code> and <code>OpaqueType</code> are gone,
1151 and all APIs deal with <code>Type*</code> instead of <code>const
1152 Type*</code>. If you need to create recursive structures, then create a
1153 named structure, and use <code>setBody()</code> when all its elements are
1154 built. Type merging and refining is gone too: named structures are not
1155 merged with other structures, even if their layout is identical. (of
1156 course anonymous structures are still uniqued by layout).</li>
1158 <li><code>PHINode::reserveOperandSpace</code> has been removed. Instead, you
1159 must specify how many operands to reserve space for when you create the
1160 PHINode, by passing an extra argument
1161 into <code>PHINode::Create</code>.</li>
1163 <li>PHINodes no longer store their incoming BasicBlocks as operands. Instead,
1164 the list of incoming BasicBlocks is stored separately, and can be accessed
1165 with new functions <code>PHINode::block_begin</code>
1166 and <code>PHINode::block_end</code>.</li>
1168 <li>Various functions now take an <code>ArrayRef</code> instead of either a
1169 pair of pointers (or iterators) to the beginning and end of a range, or a
1170 pointer and a length. Others now return an <code>ArrayRef</code> instead
1171 of a reference to a <code>SmallVector</code>
1172 or <code>std::vector</code>. These include:
1174 <!-- Please keep this list sorted. -->
1175 <li><code>CallInst::Create</code></li>
1176 <li><code>ComputeLinearIndex</code> (in <code>llvm/CodeGen/Analysis.h</code>)</li>
1177 <li><code>ConstantArray::get</code></li>
1178 <li><code>ConstantExpr::getExtractElement</code></li>
1179 <li><code>ConstantExpr::getGetElementPtr</code></li>
1180 <li><code>ConstantExpr::getInBoundsGetElementPtr</code></li>
1181 <li><code>ConstantExpr::getIndices</code></li>
1182 <li><code>ConstantExpr::getInsertElement</code></li>
1183 <li><code>ConstantExpr::getWithOperands</code></li>
1184 <li><code>ConstantFoldCall</code> (in <code>llvm/Analysis/ConstantFolding.h</code>)</li>
1185 <li><code>ConstantFoldInstOperands</code> (in <code>llvm/Analysis/ConstantFolding.h</code>)</li>
1186 <li><code>ConstantVector::get</code></li>
1187 <li><code>DIBuilder::createComplexVariable</code></li>
1188 <li><code>DIBuilder::getOrCreateArray</code></li>
1189 <li><code>ExtractValueInst::Create</code></li>
1190 <li><code>ExtractValueInst::getIndexedType</code></li>
1191 <li><code>ExtractValueInst::getIndices</code></li>
1192 <li><code>FindInsertedValue</code> (in <code>llvm/Analysis/ValueTracking.h</code>)</li>
1193 <li><code>gep_type_begin</code> (in <code>llvm/Support/GetElementPtrTypeIterator.h</code>)</li>
1194 <li><code>gep_type_end</code> (in <code>llvm/Support/GetElementPtrTypeIterator.h</code>)</li>
1195 <li><code>GetElementPtrInst::Create</code></li>
1196 <li><code>GetElementPtrInst::CreateInBounds</code></li>
1197 <li><code>GetElementPtrInst::getIndexedType</code></li>
1198 <li><code>InsertValueInst::Create</code></li>
1199 <li><code>InsertValueInst::getIndices</code></li>
1200 <li><code>InvokeInst::Create</code></li>
1201 <li><code>IRBuilder::CreateCall</code></li>
1202 <li><code>IRBuilder::CreateExtractValue</code></li>
1203 <li><code>IRBuilder::CreateGEP</code></li>
1204 <li><code>IRBuilder::CreateInBoundsGEP</code></li>
1205 <li><code>IRBuilder::CreateInsertValue</code></li>
1206 <li><code>IRBuilder::CreateInvoke</code></li>
1207 <li><code>MDNode::get</code></li>
1208 <li><code>MDNode::getIfExists</code></li>
1209 <li><code>MDNode::getTemporary</code></li>
1210 <li><code>MDNode::getWhenValsUnresolved</code></li>
1211 <li><code>SimplifyGEPInst</code> (in <code>llvm/Analysis/InstructionSimplify.h</code>)</li>
1212 <li><code>TargetData::getIndexedOffset</code></li>
1215 <li>All forms of <code>StringMap::getOrCreateValue</code> have been remove
1216 except for the one which takes a <code>StringRef</code>.</li>
1218 <li>The <code>LLVMBuildUnwind</code> function from the C API was removed. The
1219 LLVM <code>unwind</code> instruction has been deprecated for a long time
1220 and isn't used by the current front-ends. So this was removed during the
1221 exception handling rewrite.</li>
1223 <li>The <code>LLVMAddLowerSetJmpPass</code> function from the C API was
1224 removed because the <code>LowerSetJmp</code> pass was removed.</li>
1226 <li>The <code>DIBuilder</code> interface used by front ends to encode
1227 debugging information in the LLVM IR now expects clients to
1228 use <code>DIBuilder::finalize()</code> at the end of translation unit to
1229 complete debugging information encoding.</li>
1231 <li>TargetSelect.h moved to Support/ from Target/</li>
1233 <li>UpgradeIntrinsicCall no longer upgrades pre-2.9 intrinsic calls (for
1234 example <code>llvm.memset.i32</code>).</li>
1236 <li>It is mandatory to initialize all out-of-tree passes too and their dependencies now with
1237 <code>INITIALIZE_PASS{BEGIN,END,}</code>
1238 and <code>INITIALIZE_{PASS,AG}_DEPENDENCY</code>.</li>
1240 <li>The interface for MemDepResult in MemoryDependenceAnalysis has been
1241 enhanced with new return types Unknown and NonFuncLocal, in addition to
1242 the existing types Clobber, Def, and NonLocal.</li>
1249 <!-- *********************************************************************** -->
1251 <a name="knownproblems">Known Problems</a>
1253 <!-- *********************************************************************** -->
1257 <p>LLVM is generally a production quality compiler, and is used by a broad range
1258 of applications and shipping in many products. That said, not every
1259 subsystem is as mature as the aggregate, particularly the more obscure
1260 targets. If you run into a problem, please check the <a
1261 href="http://llvm.org/bugs/">LLVM bug database</a> and submit a bug if
1262 there isn't already one or ask on the <a
1263 href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVMdev
1266 <p>Known problem areas include:</p>
1269 <li>The Alpha, Blackfin, CellSPU, MSP430, PTX, SystemZ and
1270 XCore backends are experimental, and the Alpha, Blackfin and SystemZ
1271 targets have already been removed from mainline.</li>
1273 <li>The integrated assembler, disassembler, and JIT is not supported by
1274 several targets. If an integrated assembler is not supported, then a
1275 system assembler is required. For more details, see the <a
1276 href="CodeGenerator.html#targetfeatures">Target Features Matrix</a>.
1279 <li>The C backend has numerous problems and is not being actively maintained.
1280 Depending on it for anything serious is not advised.</li>
1287 <!-- *********************************************************************** -->
1289 <a name="additionalinfo">Additional Information</a>
1291 <!-- *********************************************************************** -->
1295 <p>A wide variety of additional information is available on
1296 the <a href="http://llvm.org/">LLVM web page</a>, in particular in
1297 the <a href="http://llvm.org/docs/">documentation</a> section. The web page
1298 also contains versions of the API documentation which is up-to-date with the
1299 Subversion version of the source code. You can access versions of these
1300 documents specific to this release by going into the "<tt>llvm/doc/</tt>"
1301 directory in the LLVM tree.</p>
1303 <p>If you have any questions or comments about LLVM, please feel free to contact
1304 us via the <a href="http://llvm.org/docs/#maillist"> mailing lists</a>.</p>
1308 <!--=========================================================================-->
1310 <!-- EH details: to be moved to a blog post:
1315 <p>One of the biggest changes is that 3.0 has a new exception handling
1316 system. The old system used LLVM intrinsics to convey the exception handling
1317 information to the code generator. It worked in most cases, but not
1318 all. Inlining was especially difficult to get right. Also, the intrinsics
1319 could be moved away from the <code>invoke</code> instruction, making it hard
1320 to recover that information.</p>
1322 <p>The new EH system makes exception handling a first-class member of the IR. It
1323 adds two new instructions:</p>
1326 <li><a href="LangRef.html#i_landingpad"><code>landingpad</code></a> —
1327 this instruction defines a landing pad basic block. It contains all of the
1328 information that's needed by the code generator. It's also required to be
1329 the first non-PHI instruction in the landing pad. In addition, a landing
1330 pad may be jumped to only by the unwind edge of an <code>invoke</code>
1333 <li><a href="LangRef.html#i_resume"><code>resume</code></a> — this
1334 instruction causes the current exception to resume traveling up the
1335 stack. It replaces the <code>@llvm.eh.resume</code> intrinsic.</li>
1338 <p>Converting from the old EH API to the new EH API is rather simple, because a
1339 lot of complexity has been removed. The two intrinsics,
1340 <code>@llvm.eh.exception</code> and <code>@llvm.eh.selector</code> have been
1341 superseded by the <code>landingpad</code> instruction. Instead of generating
1342 a call to <code>@llvm.eh.exception</code> and <code>@llvm.eh.selector</code>:
1344 <div class="doc_code">
1346 Function *ExcIntr = Intrinsic::getDeclaration(TheModule,
1347 Intrinsic::eh_exception);
1348 Function *SlctrIntr = Intrinsic::getDeclaration(TheModule,
1349 Intrinsic::eh_selector);
1351 // The exception pointer.
1352 Value *ExnPtr = Builder.CreateCall(ExcIntr, "exc_ptr");
1354 std::vector<Value*> Args;
1355 Args.push_back(ExnPtr);
1356 Args.push_back(Builder.CreateBitCast(Personality,
1357 Type::getInt8PtrTy(Context)));
1359 <i>// Add selector clauses to Args.</i>
1361 // The selector call.
1362 Builder.CreateCall(SlctrIntr, Args, "exc_sel");
1366 <p>You should instead generate a <code>landingpad</code> instruction, that
1367 returns an exception object and selector value:</p>
1369 <div class="doc_code">
1371 LandingPadInst *LPadInst =
1372 Builder.CreateLandingPad(StructType::get(Int8PtrTy, Int32Ty, NULL),
1375 Value *LPadExn = Builder.CreateExtractValue(LPadInst, 0);
1376 Builder.CreateStore(LPadExn, getExceptionSlot());
1378 Value *LPadSel = Builder.CreateExtractValue(LPadInst, 1);
1379 Builder.CreateStore(LPadSel, getEHSelectorSlot());
1383 <p>It's now trivial to add the individual clauses to the <code>landingpad</code>
1386 <div class="doc_code">
1388 <i><b>// Adding a catch clause</b></i>
1389 Constant *TypeInfo = getTypeInfo();
1390 LPadInst->addClause(TypeInfo);
1392 <i><b>// Adding a C++ catch-all</b></i>
1393 LPadInst->addClause(Constant::getNullValue(Builder.getInt8PtrTy()));
1395 <i><b>// Adding a cleanup</b></i>
1396 LPadInst->setCleanup(true);
1398 <i><b>// Adding a filter clause</b></i>
1399 std::vector<Constant*> TypeInfos;
1400 Constant *TypeInfo = getFilterTypeInfo();
1401 TypeInfos.push_back(Builder.CreateBitCast(TypeInfo, Builder.getInt8PtrTy()));
1403 ArrayType *FilterTy = ArrayType::get(Int8PtrTy, TypeInfos.size());
1404 LPadInst->addClause(ConstantArray::get(FilterTy, TypeInfos));
1408 <p>Converting from using the <code>@llvm.eh.resume</code> intrinsic to
1409 the <code>resume</code> instruction is trivial. It takes the exception
1410 pointer and exception selector values returned by
1411 the <code>landingpad</code> instruction:</p>
1413 <div class="doc_code">
1415 Type *UnwindDataTy = StructType::get(Builder.getInt8PtrTy(),
1416 Builder.getInt32Ty(), NULL);
1417 Value *UnwindData = UndefValue::get(UnwindDataTy);
1418 Value *ExcPtr = Builder.CreateLoad(getExceptionObjSlot());
1419 Value *ExcSel = Builder.CreateLoad(getExceptionSelSlot());
1420 UnwindData = Builder.CreateInsertValue(UnwindData, ExcPtr, 0, "exc_ptr");
1421 UnwindData = Builder.CreateInsertValue(UnwindData, ExcSel, 1, "exc_sel");
1422 Builder.CreateResume(UnwindData);
1432 <!-- *********************************************************************** -->
1436 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1437 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
1438 <a href="http://validator.w3.org/check/referer"><img
1439 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
1441 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
1442 Last modified: $Date$